There is a lot of talk about "long-horizon agents, " AI that continues to work on a task over an extended period of time.
This, to me, seems no different than the long-running jobs that engineers are already familiar with, and the pitfalls they can bring.
Yep, spot on! I'm thinking the same way, to the extent that in Inngest runs, I specify a separate failure function that evaluates and cleans up when necessary.
Thanks for your take on this, i'm wanting to understand patterns that are emerging with longer running processes/agents!
@Tech_p001 Getting debug flashbacks now 😅 That's why I prefer modern event-based workflow orchestration these days, much looser coupling with the same benefits.
I'm biased toward Inngest, but BullMQ and Temporal would work as well.
@bradey@jason_cxyz@luizfper@ThePrimeagen I use Debian/Ubuntu a ton managing servers, that is my goto option, but Omarchy sounds cool! I might have to spin it up on my trash bench machine just to try it. 👀
Right on, do you think that both of these problems could be solved by checkpointing? Or is the likelyhood of the state not being fresh a problem that can't live between retries?
Partial work is an interesting one; do we just retry, or do we need rollback procedures to "undo" work?
@luizfper@ThePrimeagen I considered the Framework, but parts of it (the screen) still look like a laptop from 2015; I also heard that the keyboard flexes a lot, which would drive me crazy!
We've been secretly working on a super cool project at @Inngest for the past few weeks, and it launches today! Teaming up with @codetv_dev, we challenged a group of brilliant engineers to build an app that does work while you're away.
https://t.co/Y5OCvFFxlI
I'll be giving a lightning talk at @AIconference 2026 in San Francisco, Sept 29 – Oct 1: 120+ speakers, 5 tracks, 5,500+ AI builders.
Use code "NETWORK25" for a discount: https://t.co/EjRJaYjDRr #theaiconference#taic2026
@andrew_n_carr I’ve been on LinkedIn for a long time; got quite a bit of work from it; but I’ve heard many times now that X is where its at for building a tech audience, any tips? 😅
@v0xium I felt the exact same way at my previous job. I built the platform by hand pre-ai; an extremely high throughput data ingestion system with finely tuned memory management- just to be run ragged by leadership as soon as AI hit the scene.